Neural Networks - Science topic

Why do Long Short-Term Memory (LSTM) networks generally exhibit lower Mean Squared Error (MSE) compared to traditional Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) in certain applications?

https://youtu.be/VQDB6uyd_5E In this video, we explore why Long Short-Term Memory (LSTM) networks often achieve lower Mean Squared Error (MSE) compared to traditional Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) in specific applications. We delve into the unique architecture of LSTMs, their ability to handle long-range dependencies, and how they mitigate issues like the vanishing gradient problem, leading to improved performance in tasks such as sequence modeling and time series prediction. Topics Covered: 1. Understanding the architecture and mechanisms of LSTMs 2. Comparison of LSTM, RNN, and CNN in terms of MSE performance 3. Handling long-range dependencies and vanishing gradients 4. Applications where LSTMs excel and outperform traditional neural networks Watch this video to discover why LSTMs are favored for certain applications and how they contribute to lower MSE in neural network models! #LSTM #RNN #CNN #NeuralNetworks #DeepLearning #MachineLearning #MeanSquaredError #SequenceModeling #TimeSeriesPrediction #VanishingGradient #AI Don't forget to like, comment, and subscribe for more content on neural networks, deep learning, and machine learning concepts! Let's dive into the world of LSTMs and their impact on model performance. Feedback link: https://maps.app.goo.gl/UBkzhNi7864c9BB1A LinkedIn link for professional queries: https://www.linkedin.com/in/professorrahuljain/ Join my Telegram link for Free PDFs: https://t.me/+xWxqVU1VRRwwMWU9 Connect with me on Facebook: https://www.facebook.com/professorrahuljain/ Watch Videos: Professor Rahul Jain  Link: https://www.youtube.com/@professorrahuljain

2024 4th International Conference on Computer, Remote Sensing and Aerospace (CRSA 2024) will be held at Osaka, Japan on July 5-7, 2024.

The topics of interest for submission include, but are not limited to:

1. Algorithms

Image Processing

Data processing

Data Mining

Computer Vision

Computer Aided Design

......

2. Remote Sensing

Optical Remote Sensing

Microwave Remote Sensing

Remote Sensing Information Engineering

Geographic Information System

Global Navigation Satellite System

......

3. Aeroacoustics

Aeroelasticity and structural dynamics

Aerothermodynamics

Airworthiness

Autonomy

Mechanisms

......

All accepted papers will be published in the Conference Proceedings, and submitted to EI Compendex, Scopus for indexing.

Full Paper Submission Date: May 31, 2024

Registration Deadline: May 31, 2024

Conference Date: July 5-7, 2024

Invitation code: AISCONF

*Using the invitation code on submission system/registration can get priority review and feedback

Given a multi-layer (say 10-12) neural network, are there standard techniques to compress it to a single layer or 2 layer NN ?

What are the most effective techniques for mitigating overfitting in neural networks, especially when dealing with limited training data?

memorize-ability > generalize-ability

Hello everyone and thank you for reading my question.

I have a data set that have around 2000 data point. It have 5 inputs (4 wells rate and the 5th is the time) and 2 ouputs ( oil cumulative and water cumulative). See the attached image.

I want to build a Proxy model to simualte the cumulative oil & water.

I have made 5 models ( ANN, Extrem Gradient Boost, Gradient Boost, Randam forest, SVM) and i have used GridSearch to tune the hyper parameters and the results for training the models are good. Of course I have spilited the training data set to training, test and validation sets.

So I have another data that I haven't include in either of the train,test and validation sets and when I use the models to predict the output for this data set the models results are bad ( failed to predict).

I think the problem lies in the data itself because the only input parameter that changes are the (days) parameter while the other remains constant.

But the problem is I can't remove the well rate or join them into a single variable because after the Proxy model has been made I want to optimize the well rates to maximize oil and minimize water cumulative respectively.

Is there a solution to suchlike issue?

Chalmers in his book: What is this thing called Science? mentions that Science is Knowledge obtained from information. The most important endeavors of science are : Prediction and Explanation of Phenomenon. The emergence of Big (massive) Data leads us to the field of Data Science (DS) with the main focus on prediction. Indeed, data belong to a specific field of knowledge or science (physics, economy, ....).

If DS is able to realize prediction for the field of sociology (for example), to whom the merit is given: Data Scientist or Sociologist?

10.1007/s11229-022-03933-2

#DataScience #ArtificialIntelligence #Naturallanguageprocessing #DeepLearning #Machinelearning #Science #Datamining

How do you become a Machine Learning(ML) and Artificial Intelligence(AI) Engineer? or start research in AI/ML, Neural Networks, and Deep Learning?

Should I pursue a "Master of Science thesis in Computer Science." with a major in AI to become an AI Engineer?

I am researching on automatic modulation classification (AMC). I used the "RADIOML 2018.01A" dataset to simulate AMC and used the convolutional long-short term deep neural network (CLDNN) method to model the neural network. But now I want to generate the dataset myself in MATLAB.

My question is, do you know a good sources (papers or codes) that have produced a dataset for AMC in MATLAB (or Python)? In fact, have they produced the In-phase and Quadrature components for different modulations (preferably APSK and PSK)?

How does the addition of XAI techniques such as SHAP or LIME impact model interpretability in complex machine learning models like deep neural networks?

Perhaps in the future - as a result of the rapid technological advances currently taking place and the rivalry of leading technology companies developing AI technologies - a general artificial intelligence (AGI) will emerge. At present, there are unresolved deliberations on the question of new opportunities and threats that may occur as a result of the construction and development of general artificial intelligence in the future. The rapid technological progress currently taking place in the field of generative artificial intelligence in connection with the already high level of competition among technology companies developing these technologies may lead to the emergence of a super artificial intelligence, a strong general artificial intelligence that can achieve the capabilities of self-development, self-improvement and perhaps also autonomy, independence from humans. This kind of scenario may lead to a situation where this kind of strong, super AI or general artificial intelligence is out of human control. Perhaps this kind of strong, super, general artificial intelligence will be able, as a result of self-improvement, to reach a state that can be called artificial consciousness. On the one hand, new possibilities can be associated with the emergence of this kind of strong, super, general artificial intelligence, including perhaps new possibilities for solving the key problems of the development of human civilization. However, on the other hand, one should not forget about the potential dangers if this kind of strong, super, general artificial intelligence in its autonomous development and self-improvement independent of man were to get completely out of the control of man. Probably, whether this will involve mainly new opportunities or rather new dangers for mankind will mainly be determined by how man will direct this development of AI technology while he still has control over this development.

I described the key issues of opportunities and threats to the development of artificial intelligence technology in my article below:

In view of the above, I address the following question to the esteemed community of scientists and researchers:

What do you think about this topic?

What is your opinion on this issue?

Please answer,

I invite everyone to join the discussion,

Thank you very much,

Best regards,

Dariusz Prokopowicz

In other words, why have improvements to neural networks led to an increase in hyperparameters? Are hyperparameters related to some fundamental flaw of neural networks?

What approaches can be used to enhance the interpretability of deep neural networks for better understanding of their decision-making process ?

I am looking for a Q1 journal with a publication cost of 0 USD and a very short publishing period, specifically in the field of Hybrid Neural Networks. Can anyone suggest some?

Thank you.

I would like to know that prophet time series model is under the category of neural network or machine learning or deep learning? I want to forecast the price of product depending on other influential factors( 7 indicators) and all the data is monthly data with 15 years period.How can I implement with prophet model to get better accuracy? And i also want to compare the result with other time series model.Please suggest me how should I do about my work.thank you.

I have trained all Convolutional Neural Networks (CNNs) from the LeNet-5 model to the EfficientNet model for Benign Tumors and Malignant Tumors for breast masses with large data. The data was Mammogram Images(MI). All of these models give me a test accuracy of 50 %. Why did most journals publish fake results?

The future of AI holds boundless potential across various domains, poised to transform industries, societies, and everyday lives. Advancements in machine learning, deep learning, and neural networks continue to push the boundaries of what AI can achieve.

We anticipate AI systems becoming increasingly integrated into our daily routines, facilitating more personalized experiences in healthcare, education, entertainment, and beyond.

Collaborative efforts between technologists, policymakers, and ethicists will be essential to ensure AI development remains aligned with human values and societal well-being.

As AI algorithms become more sophisticated, they will enhance decision-making processes, optimize resource allocation, and drive innovation across sectors.

However, the future of AI also raises ethical, privacy, and employment concerns that necessitate careful consideration and regulation.

As AI evolves, fostering transparency, accountability, and inclusivity will be imperative to harness its transformative potential responsibly and equitably, shaping a future where AI serves as a powerful tool for positive change.

Does Neural Networks handle a multicollinearity?

I heard about ARTIFICIAL NEURAL NETWORK (ANN) and I watched a video of a researcher talked about this revolution.. However, is ANN will be the next solution to predict the adsorption behaviour and do the adsorption calculations based on the properties of the adsorbent materials?

My paper "Bringing uncertainty quantification to the extreme-edge with memristor-based Bayesian neural networks" has been published in nature communication since the 20th November. But on google scholar, only the pre-print from research square is available...

..

Data is part of the code.

Neural network is actually code for fuzzy match.

If an activation function has a jump discontinuity, then in the training process, can we implement backpropagation to compute the derivatives and update the parameters?

..

In the rapidly evolving landscape of the Internet of Things (IoT), the integration of blockchain, machine learning, and natural language processing (NLP) holds promise for strengthening cybersecurity measures. This question explores the potential synergies among these technologies in detecting anomalies, ensuring data integrity, and fortifying the security of interconnected devices.

..

..

..

Imagine training a neural network on data like weather patterns, notoriously chaotic and unpredictable. Can the network, without any hints or constraints, learn to identify and repeat hidden periodicities within this randomness? This question explores the possibility of neural networks spontaneously discovering order in chaos, potentially revealing new insights into complex systems and their modeling through AI.

Imagine machines that can think and learn like humans! That's what AI is all about. It's like teaching computers to be smart and think for themselves. They can learn from mistakes, understand what we say, and even figure things out without being told exactly what to do.

Just like a smart friend helps you, AI helps machines be smart too. It lets them use their brains to understand what's going on, adjust to new situations, and even solve problems on their own. This means robots can do all sorts of cool things, like helping us at home, driving cars, or even playing games!

There's so much happening in Artificial Intelligence (AI), with all sorts of amazing things being developed for different areas. So, let's discuss all the cool stuff AI is being used for and the different ways it's impacting our lives. From robots and healthcare to art and entertainment, anything and everything AI is up to is on the table!

Machine Learning: Computers can learn from data and improve their performance over time, like a student studying for a test.

Natural Language Processing (NLP): AI can understand and generate human language, like a translator who speaks multiple languages.

Computer Vision: Machines can interpret and make decisions based on visual data, like a doctor looking at an X-ray.

Robotics: AI helps robots perceive their environment and make decisions, like a self-driving car navigating a busy street.

Neural Networks: Artificial neural networks are inspired by the human brain and are used in many AI applications, like a chess computer that learns to make winning moves.

Ethical AI: We need to use AI responsibly and address issues like bias, privacy, and job displacement, like making sure a hiring algorithm doesn't discriminate against certain groups of people.

Autonomous Vehicles: AI-powered cars can drive themselves, like a cruise control system that can take over on long highway drives.

AI in Healthcare: AI can help doctors diagnose diseases, plan treatments, and discover new drugs, like a virtual assistant that can remind patients to take their medication.

Virtual Assistants: AI-powered virtual assistants like Siri, Alexa, and Google Assistant can understand and respond to human voice commands, like setting an alarm or playing music.

Game AI: AI is used in games to create intelligent and challenging enemies and make the game more fun, like a boss battle in a video game that gets harder as you play.

Deep Learning: Deep learning is a powerful type of machine learning used for complex tasks like image and speech recognition, like a self-driving car that can recognize stop signs and traffic lights.

Explainable AI (XAI): As AI gets more complex, we need to understand how it makes decisions to make sure it's fair and unbiased, like being able to explain why a loan application was rejected.

Generative AI: AI can create new content like images, music, and even code, like a program that can write poetry or compose music.

AI in Finance: AI is used in the financial industry for things like algorithmic trading, fraud detection, and customer service, like a system that can spot suspicious activity on a credit card.

Smart Cities: AI can help make cities more efficient and sustainable, like using traffic cameras to reduce congestion.

Facial Recognition: AI can be used to recognize people's faces, but there are concerns about privacy and misuse, like using facial recognition to track people without their consent.

AI in Education: AI can be used to personalize learning, automate tasks, and provide educational support, like a program that can tutor students in math or English.

..

..

This question blends various emerging technologies to spark discussion. It asks if sophisticated image recognition AI, trained on leaked bioinformatics data (e.g., genetic profiles), could identify vulnerabilities in medical devices connected to the Internet of Things (IoT). These vulnerabilities could then be exploited through "quantum-resistant backdoors" – hidden flaws that remain secure even against potential future advances in quantum computing. This scenario raises concerns for cybersecurity, ethical hacking practices, and the responsible development of both AI and medical technology.

..

..

..

---𝐂𝐚𝐥𝐥 𝐅𝐨𝐫 𝐏𝐚𝐩𝐞𝐫𝐬---

The topics of interest for submission include, but are not limited to:

- Neural Networks

- Signal and information processing

- Integrated Circuit Engineering

- Electronic and Communication Engineering

- Communication and Information System

All accepted papers will be published in IEEE（ISBN:979-8-3503-9437-5）, which will be submitted for indexing by IEEE Xplore, Ei Compendex, Scopus.

𝐈𝐦𝐩𝐨𝐫𝐭𝐚𝐧𝐭 𝐃𝐚𝐭𝐞𝐬:

Full Paper Submission Date: November 12, 2023

Registration Deadline: November 28, 2023

Final Paper Submission Date: December 22, 2023

Conference Dates: January 19-21, 2024

𝐅𝐨𝐫 𝐌𝐨𝐫𝐞 𝐃𝐞𝐭𝐚𝐢𝐥𝐬 𝐩𝐥𝐞𝐚𝐬𝐞 𝐯𝐢𝐬𝐢𝐭:

..

..

..

..

..

This question delves into the domain of deep learning, focusing on regularization techniques. Regularization helps prevent overfitting in neural networks, but this question specifically addresses methods aimed at improving interpretability while maintaining high performance. Interpretability is crucial for understanding and trusting complex models, especially in fields like healthcare or finance. The question invites exploration into innovative and lesser-known techniques designed for this nuanced balance between model performance and interpretability.

..

..

..

What datasets other than ImageNet, CIFAR-10 or CINIC-10 can be used to train a simple neural network?

..

Finding the best features (sometimes called metrics or parameters) to input into a neural network by trial and error can be a very lengthy process. Classic feature selection methods in machine learning are 'extra tree classifiers', 'univariate feature selection', 'recursive feature elimination', and 'linear discrimination analysis' (a supervised learning version of PCA). Are there other more modern methods that have evolved recently which are more powerful than these?

Inputting too many redundant or worthless features into a neural network reduces the accuracy, as does omitting the most useful features. Restricting the neural network input those most relavant features is key to getting the highest accuracy from a neural network.

The decision-making process in Neural Networks poses a significant challenge known as the 'Black Box' problem. NNs are compelling in various applications, but issues could arise when accountability becomes crucial. How can one address the challenge of ensuring that a model is free from decision-making biases, and to what extent does this challenge affect the entire industry? Are there any papers or books that delve into the 'Black Box' problem and provide insights into ensuring that NNs make unbiased decisions?

I'm aware of the gradient descent and the back-propagation algorithm. What I don't get is: when is using a bias important and how do you use it?

Iam planning to do some literature work on rationale neural networks and functionalities of activation functions like sigmoid and others, please recommend me some effective articles related one

I am preparing my Bachelor final thesis in computer engineering. I am currently planning out the work. My idea is to compare traditional approaches to building recommender systems to Graph Neural Network based approaches. The plan so far is to use the Movie Lens 100k dataset, which contains data on users, movies, and user-movie ratings. The task of the recommender system would be to predict the missing ratings for user A and recommend movies based on that (say top 5 highest predictions). I would present three approaches to this task:

Given this very general outline, would you guys say that this seems like a good project idea? The movie lens dataset seems to be quite popular when it comes to experimenting with GNN's, but you can suggest a better dataset for this setup.

For details on the current OpenAI leadership situation see e.g.

https://www.theguardian.com/technology/2023/nov/19/openai-investors-push-for-return-of-ousted-ceo-sam-altman-chatgpt

Almost from the very beginning of the development of ICT, the rivalry between IT professionals operating on two sides of the barricade, i.e. in the sphere of cybercrime and cyber security, has been realized. In a situation where, within the framework of the technological progress that is taking place, on the one hand, a new technology emerges that facilitates the development of remote communication, digital transfer and processing of data then, on the other hand, the new technology is also used within the framework of hacking and/or cybercrime activities. Similarly, when the Internet appeared then on the one hand a new sphere of remote communication and digital data transfer was created. On the other hand, new techniques of hacking and cybercriminal activities were created, for which the Internet became a kind of perfect environment for development. Now, perhaps, the next stage of technological progress is taking place, consisting of the transition of the fourth into the fifth technological revolution and the development of 5.0 technology supported by the implementation of artificial neural networks based on artificial neural networks subjected to a process of deep learning constantly improved generative artificial intelligence technology. The development of generative artificial intelligence technology and its applications will significantly increase the efficiency of business processes, increase labor productivity in the manufacturing processes of companies and enterprises operating in many different sectors of the economy. Accordingly, after the implementation of generative artificial intelligence and also Big Data Analytics and other technologies typical of the current fourth technological revolution, the competition between IT professionals operating on two sides of the barricade, i.e., in the sphere of cybercrime and cybersecurity, will probably change. However, what will be the essence of these changes?

In view of the above, I address the following question to the esteemed community of scientists and researchers:

What do you think about this topic?

What is your opinion on this issue?

Please answer,

I invite everyone to join the discussion,

Thank you very much,

Best regards,

Dariusz Prokopowicz

Can you explain the concept of the vanishing gradient problem in deep learning? How does it affect the training of deep neural networks, and what techniques or architectures have been developed to mitigate this issue?

Solutions to this question may vary. However, the key issue is the moral dilemmas in the applications of the constantly developing and improving artificial intelligence technology and the preservation of ethics in the process of developing applications of these technologies. In addition to this, the key issues within the framework of this issue also include the need to more fully explore and clarify what human consciousness is, how it is formed, how it functions within specific plexuses of neurons in the human central nervous system.

In view of the above, I address the following question to the esteemed community of scientists and researchers:

And what is your opinion on this topic?

What do you think about this topic?

Please answer,

I invite everyone to join the discussion,

Thank you very much,

Best wishes,

Dariusz Prokopowicz

The above text is entirely my own work written by me on the basis of my research.

In writing this text I did not use other sources or automatic text generation systems.

Copyright by Dariusz Prokopowicz

I decided to learn about such an area as the use of neural networks in econometrics, regardless of subsequent employment. One PhD explained to me that:

"In econometric research, the explainability of models is important; neural networks do not provide this. For time series, neural networks can be used, but only with a special architecture, for example, LSTM. For macroeconomic forecasting tasks, as a rule, neural networks are not used. ARIMA/SARIMA, VAR, ECM are used."

But on one forum they explained to me that

"A typical task in the field of time series analysis is to predict, from a sequence of previous values of a time series, the most likely next/future value. The Large Language Model (LLM), which underlies the same ChatGPT, predicts which word or phrase will be next in a sentence or phrase, i.e. in a sequence of words in natural language. The current ChatGPT is implemented using so-called transformers - neural networks, which after 2017 began to actively replace the older, but also neural network and also sequence-oriented LSTM (long short-term memory networks) architecture, and not only in text processing tasks, but also in other areas."

That is, the use of transformers in time series forecasting may seem promising? It seems that now this is a relatively young industry, still little studied?

Dear Expert,

I used the Neural Network in MATLAB using inputs [10*3] target data [10*1] and one Hidden layer [25 neurons]. then How can I create an equation that correctly estimates the predicted target??

(Based on the ANN created, weights, biases, and related inputs)

Is there a method, tool, or idea to solve this issue and create one final equation that predicts the output?

I am new to machine learning, I working on regression neural network to prediction the outcomes of my experiments. I created the neural network with a hidden layer to predict my outcomes, now i have to tune the hyperparameter to optimize the NN.

Could you elaborate on the difficulties and obstacles that arise when training deep neural networks, and how researchers and practitioners have attempted to address these challenges?

How to calculate the RMSE value especially (testing and training values) of artifical neural network by using spss ? in the output is parameter estimate heading especially output value is act as a testing and predicted value under the input layer is training ? i am attaching my parameter estimates table output for more clear understanding about it.

I constantly use genetic algorithm and neural network , if you know and examine a better method to find when the data is high dimensional .

Isn't this how humans learn? First remember some things, then make some guesses about new things based on existing memories, just like a neural network?

So, do you feel that the current path of deep learning can lead to AGI (Artificial General Intelligence)?

The intersection of neuroscience, electronics, and AI has sparked a profound debate questioning whether humanity can be considered a form of technology itself. This discourse revolves around the comparison of the human chemical-electric nodes—neurons, with the nodes of a computer, and the potential implications of transplanting human consciousness into machines.

Neurons, as the elemental building blocks of the human brain, operate through the transmission of electrochemical signals, forming a complex network that underpins cognitive functions, emotions, and consciousness. In contrast, computer nodes are physical components designed to process and transmit data through electrical signals, governed by programmed algorithms.

The notion of transferring the human mind into a machine delves into the essence of human identity and the philosophical nuances of consciousness. While it may be feasible to replicate certain cognitive functions within a machine by mimicking neural networks, there are profound ethical and philosophical implications at stake.

Critics argue that even if a machine were to replicate the intricacies of the human brain, it would lack essential human qualities such as emotions, subjective experiences, and moral reasoning, thus failing to encapsulate the essence of human consciousness. Furthermore, the concept of integrating the human mind with machines raises complex questions about the nature of identity and self-awareness. If the entirety of a human mind were to be transplanted into a machine, the resulting entity may no longer fit the traditional definition of human, but rather a hybrid of human cognition and artificial intelligence.

On the other hand, proponents of merging human minds with machines foresee the potential for significant advancements in AI and neuroscience, suggesting that through advanced brain-computer interfaces, it might be possible to enhance human cognition and expand the capabilities of the human mind, blurring the boundaries between organic and artificial intelligence.

As the realms of electronics and AI continue to evolve, the question of whether humanity itself can be perceived as a form of technology remains a deeply contemplative issue. It is imperative that as these technological frontiers advance, ethical considerations and respect for human values are prioritized, ensuring that any progression in this field aligns with the preservation of human dignity and integrity.

The advancement of technology and the intricacies involved in simulating human cognitive processes suggest that it might be plausible for machines to exhibit emotions akin to humans. As the complexity of AI systems increases, managing a vast number of nodes and intricate algorithms could potentially lead to unexpected and seemingly irrational behaviors, which might even resemble emotional responses.

Similarly to how a basic machine operates in a predictable and precise manner devoid of human characteristics, the proliferation of complexity in a machine's structure could lead to the emergence of seemingly irrational or emotional behaviors. Managing the intricate interplay between a multitude of nodes might result in the manifestation of behaviors that mimic emotions, despite the absence of genuine human experience.

These behaviors could be centered around learned and preprogrammed principles, allowing the machine to respond in a manner that mirrors human emotions.

Moreover, the ability to simulate emotions in machines has gained traction due to the growing understanding of the role of neural networks and the intricate interplay of various computational elements within AI systems. As AI models become more sophisticated, they could feasibly process information in a way that mirrors the human emotional experience, albeit based on programmed responses rather than genuine feelings.

While the debate about whether machines can truly experience emotions similar to humans remains unsettled, the increasingly complex and interconnected nature of AI systems hints at the potential for machines to display a form of emotive behavior as they grapple with the challenges of managing a multitude of nodes and algorithms.

This perspective challenges the conventional notion that emotions are exclusively tied to human consciousness and suggests that with the advancement of technology, machines might exhibit behaviors that closely resemble human emotions, albeit within the confines of programmed and learned parameters.

In the foreseeable future, it is conceivable that machines will surpass the human mind in terms of node count, compactness, and complexity, operating with heightened efficiency. As this technological advancement unfolds, it is plausible that profound questions may arise regarding whether the frequencies generated by the human brain are inferior to those generated by machines.

I now that alot of artificial networks has appeared now. And may be soon we wil not read articles and do our scientific works and AI will help us. May be it is happening now? Wat is your experience working with AI and neural networks in science?

Hello everyone! I am studying Graph Neural Networks to apply to my field.

My problem: I have a dataset with multiple graphs. Each node in a graph have Y label. I want to predict Y label of nodes in new graph.

I want to ask: I can make predictions by Graph Neural Network? If can, Could you give me some hints?

Below is a illustration about my question.

Thank you!

The experiment conducted by Bose at the Royal Society of London in 1901 demonstrated that plants have feelings like humans. Placing a plant in a vessel containing poisonous solution he showed the rapid movement of the plant which finally died down. His finding was praised and the concept of plant’s life has been established. If we scold a plant it doesn’t respond, but an AI bot does. Then how can we disprove the life of a Chatbot?

The progressive digitization of data and archived documents, digitization of data transfer processes, Internetization of communications, economic processes but also of research and analytical processes is becoming a typical feature of today's developing developed economies. Currently, another technological revolution is taking place, described as the fourth and in some aspects it is already the fifth technological revolution. Particularly rapidly developing and finding more and more applications are technologies categorized as Industry 4.0/5.0. These technologies, which support research and analytical processes carried out in various institutions and business entities, include Big Data Analytics and artificial intelligence. The computational capabilities of microprocessors, which are becoming more and more perfect and processing data faster and faster, are successively increasing. The processing of ever-larger sets of data and information is growing. Databases of data and information extracted from the Internet and processed in the course of conducting specific research and analysis processes are being created. In connection with this, the possibilities for the application of Big Data Analytics supported by artificial intelligence technology in terms of improving research techniques, in terms of increasing the efficiency of the research and analytical processes used so far, in terms of improving the scientific research being conducted, are also growing rapidly.

In view of the above, I address the following question to the esteemed community of scientists and researchers:

What do you think on this topic?

What is your opinion on this issue?

Please answer,

I invite everyone to join the discussion,

Thank you very much,

Best wishes,

The above text is entirely my own work written by me on the basis of my research.

Copyright by Dariusz Prokopowicz

On my profile of the Research Gate portal you can find several publications on Big Data issues. I invite you to scientific cooperation in this problematic area.

Dariusz Prokopowicz

I have deep neural network where I want to include a layer which should have one input and two outputs. For example, I want to construct an intermediate layer where Layer-1 is connected to the input of this intermediate layer and one output of the intermediate layer is connected to Layer-2 and another output is connected to Layer-3. Moreover, the intermediate layer just passes the data as it is through it without doing any mathematical operation on the input data. I have seen additionLayer in MATLAB, but it has only 1 output and this function is read-only for the number of outputs.

New technologies, including ICT information technologies, technologies categorized as Industry 4.0 or Industry 5.0 are finding new applications. These technologies are currently developing rapidly and are an important factor in the current fourth technological revolution. On the other hand, due to the still high emissions of greenhouse gases generating the process of global warming, due to progressive climate change, increasingly frequent weather anomalies and climatic disasters, in addition to increasing environmental pollution, still rapidly decreasing areas of forests, carried out predatory forest management, the level of biodiversity of the planet's natural ecosystems is rapidly decreasing. Therefore, it is necessary to engage new technologies, including ICT information technologies, technologies categorized as Industry 4.0/Industry 5.0, including new technologies in the field of Big Data Analytics and Artificial Intelligence in order to improve and scale up the protection of the biodiversity of the planet's natural ecosystems.

In view of the above, I address the following question to the esteemed community of scientists and researchers:

And what is your opinion on this topic?

What do you think about this topic?

Please answer,

I invite everyone to join the discussion,

Thank you very much,

Warm regards,

Dariusz Prokopowicz

The above text is entirely my own work written by me on the basis of my research.

In writing this text I did not use other sources or automatic text generation systems.

Copyright by Dariusz Prokopowicz

Need study material on Gnn

I have built a feed-forward fully connected neural network. Trying to specify its fitness function, I read a review paper by Ojha et al. (2017). The authors suggest including the accuracy of both training and test data sets in the fitness function (an evaluation metric), by which you could evaluate the performance of the neural network.

Considering that we build the neural network based on the training data set, I was wondering why we should include its accuracy (i.e., training accuracy) in the fitness function/evaluation metric. Why should the evaluation metric of parameter tuning not rely only on the test/validation accuracy?

Ref:

How can artificial intelligence break through the existing deep learning/neural network framework, and what are the directions?